Development and Uncertainty Assessment of Pedotransfer Functions for Predicting Water Contents at Specific Pressure Heads

Abstract

PTFs for water contents at specific pressure heads were developed. Covariate shift increased uncertainty in PTF predictions. Relative importance of predictors in machine learning PTFs was determined. There has been much effort to improve the performance of pedotransfer functions (PTFs) using intelligent algorithms, but the issue of covariate shift, i.e., different probability distributions in training and testing datasets, and its impact on prediction uncertainty of PTFs has been rarely addressed. The common practice in PTF generation is to randomly separate the dataset into training and testing subsets, and the outcomes of this random selection may be different if the process is subject to covariate shift. We evaluated the impact of covariate shift generated by data shuffling and detected by Kolmogorov–Smirnov test for the prediction of water contents using soil databases from Denmark and Brazil. The soil water contents at different pressure heads were predicted by developing linear and stepwise regression besides machine learning based PTFs including Gaussian process regression and ensemble method. Regression based PTFs for the Brazilian dataset resulted in better predictions compared with machine learning methods, which in their turn estimated high water contents in Danish soils more accurately. One hundred PTFs were developed for water content at specific pressure heads by data shuffling. From these, 100 sets of fitted van Genuchten parameters were obtained representing the generated uncertainty. Data shuffling led to covariate shift, resulting in uncertainty in water content prediction by the PTFs. Inherent variability of data may lead to increased prediction uncertainty. For correlated data, simple regression models performed as good as sophisticated machine learning methods. Using PTF‐predicted water contents for van Genuchten retention parameter fitting may lead to a high uncertainty.

Prediction of greenhouse gas emissions from agricultural fields with and without cover crops

“…It was observed a great symmetry exists in T soil , WC soil, , and air temperatures as their skewness values were close to zero; however, significant positive skewness for CO 2 and N 2 O fluxes indicated that mean values were greater than the median of data. The distributions of RH max and N 2 O with high values of kurtosis (>2) demonstrated the existence of high peaks around the mean and minor peaks in their long tails (Kotlar, de Jong van Lier, et al., 2019). Kurtosis and standard deviation of N 2 O for NCC scenario were slightly less than CC, showing suppressed peak and heavier shoulder of distribution.…”

“…Here, input attributes were initially mapped into a higher‐dimensional feature space using a mapping or so‐called kernel function. Then SVM found an optimal hyperplane in this multi‐dimensional space separating the mapped data with the largest margin (Kotlar, de Jong van Lier, et al., 2019). This study considered linear, polynomial, and radial basis kernel functions, respectively, called SVML, SVMP, and SVMR.…”

Comparing visible-near-infrared spectroscopy with classical regression pedotransfer functions for predicting near-saturated and saturated hydraulic conductivity of calcareous soils

“…The 102 soil samples were shuffled (Kotlar et al, 2019b) using the MATLAB software package (MathWorks, Makers of MATLAB and Simulink, 2021). Before developing the models, the shuffled data were imported to the Unscrambler X v. 9.7 software package (CAMO, Technologies Inc., 2013) and randomly divided into the calibration and validation datasets, containing nearly 75 % (76 samples) and 25 % (26 samples), respectively.…”

Updated European hydraulic pedotransfer functions with communicated uncertainties in the predicted variables (euptfv2)

“…Parameter tuning of the ranger was performed with the "caret" R package (Kuhn et al, 2017(Kuhn et al, , 2018. With the im-plemented train function, a fivefold cross-validation was repeated 10 times to tune the number of randomly selected predictor variables at each split (mtry) and find the best performing splitting rule (splitrule) during training.…”